R Notebook

library(dplyr)
read.csv("data/fbref/games_players.csv")

matchdat <- read.csv("https://projects.fivethirtyeight.com/soccer-api/club/spi_matches.csv")

## calaculating what the average performances 

homeav <- matchdat %>% filter(league_id == 2411) %>% #working out the average home xg 
                          summarise(avfor = mean(xg1, na.rm = T), avag = mean(xg2, na.rm = T)) 



homfor <- homeav[[1]] # extracting the number for home and away


homeag <- homeav[[2]]


# comparing each teams performance to average for xg for and against
home_rating_data <- matchdat %>% filter(league_id == 2411) %>%
                        group_by(team1) %>%
                        summarise(avfor = mean(xg1, na.rm = T), avag = mean(xg2, na.rm = T)) %>% 
                        mutate(xgfh = avfor - homfor, xgah = avag - homeag) %>%
                        select(team1, xgfh, xgah)

away_rating_data <- matchdat %>% filter(league_id == 2411) %>%
                        group_by(team2) %>%
                        summarise(avfor = mean(xg1, na.rm = T), avag = mean(xg2, na.rm = T)) %>% 
                        mutate(xgfa = avfor - homfor, xgaa = avag - homeag) %>%
                        select(team2, xgfa, xgaa)

## creating a data frame with matches the for and against scores and the xg deltas 
matches <- matchdat %>% filter(league_id == 2411) %>%
                              left_join(home_rating_data, by = "team1") %>%
                              left_join(away_rating_data, by = "team2") %>%
                              mutate(deltafh =  xgfh + xgaa, deltagh = xgah+xgfa) %>%
                              select(season, score1, score2, deltafh, deltagh)

matches

## splitting it up for home and away

mat1 <- matches %>% select(season, score1, deltafh) %>%
                      mutate(loc = "home")
colnames(mat1)[2] <- "score"
colnames(mat1)[3] <- "delta"
mat2 <- matches %>% select(season, score2, deltagh) %>%
                      mutate(loc = "away")
colnames(mat2)[2] <- "score"
colnames(mat2)[3] <- "delta"


## putting it together and creating the score as a factor
matall <- mat1 %>% bind_rows(mat2) %>%
                   mutate(scorecat = as.factor(if_else(score > 5,"5", as.character(score)))) %>% ### rating from 0 to 5. 5 being biggest score
                   select(-score, -season) ### data for calculating the chance of goals scored 

matall

library(tidymodels)

##splitting the data into training and testing
score_split <- initial_split(matall, prop = 0.9, strata = scorecat) 
score_train <- training(score_split)
score_test <- testing(score_split)

## creating the classification random forest
rand1 <- rand_forest() %>% #type of model 
                set_engine("randomForest") %>% #  engine used to fit the model
                  set_mode("classification") %>%
                  fit(scorecat ~., data = score_train)

rand1

parsnip model object


Call:
 randomForest(x = maybe_data_frame(x), y = y) 
               Type of random forest: classification
                     Number of trees: 500
No. of variables tried at each split: 1

        OOB estimate of  error rate: 65.83%
Confusion matrix:
    0    1 2 3 4 5 class.error
0 230  934 0 0 0 0   0.8024055
1 169 1172 0 0 0 0   0.1260254
2 117  775 0 0 0 0   1.0000000
3  61  386 0 0 0 0   1.0000000
4  21  148 0 0 0 0   1.0000000
5   9   81 0 0 0 0   1.0000000

### filtering the matches to test the model on 

matches_test <- matchdat %>% filter(league_id == 2411) %>%
                              left_join(home_rating_data, by = "team1") %>%
                               left_join(away_rating_data, by = "team2") %>%
                                  mutate(deltafh =  xgfh + xgaa, deltagh = xgah+xgfa) %>%
                                    filter(season == 2020)

match20_h <- matches_test %>% select(deltafh, score1) %>% 
                                mutate(loc = "home")

match20_a <- matches_test %>% select(deltagh, score2) %>%
                                mutate(loc = "away")

colnames(match20_h)[1] <- "delta"
colnames(match20_h)[2] <- "score"

colnames(match20_a)[1] <- "delta"
colnames(match20_a)[2] <- "score"

match20 <- match20_h %>% bind_rows(match20_a)

match20 <- match20 %>% bind_rows(match20) %>%
                   mutate(scorecat = as.factor(if_else(score > 5,"5", as.character(score)))) %>% ### rating from 0 to 5. 5 being biggest score
                   select(-score) ### data for calculating the chance of goals scored 

match20

library(caret)
score_test$scorecat = as.factor(score_test$scorecat)
x <- predict(rand1, score_test, type = "prob")

Error in predict.randomForest(object = object$fit, newdata = new_data,  : 
  New factor levels not present in the training data

library(fplscrapR)
fix20 <- get_game_list(season = 20)

season18 <- get_player_details(season = 18)
season19 <- get_player_details(season = 19)
season20 <- get_player_details(season = 20)

season20 %>% relocate(playername, .before = fixture)

read.csv("data/fbref/games_players.csv")

NA